一种基于全局-局部双分支注意力网络的绝缘子缺陷检测算法

兰州理工大学学报 ›› 2025, Vol. 51 ›› Issue (6): 107-115.

• 自动化技术与计算机技术 • 上一篇下一篇

一种基于全局-局部双分支注意力网络的绝缘子缺陷检测算法

焦亮¹, 王宗顺², 王思润¹, 李策^*3

1.天水长城开关厂集团有限公司, 甘肃天水 741020;
2.兰州理工大学自动化与电气工程学院, 甘肃兰州 730050;
3.兰州理工大学微电子现代产业学院, 甘肃兰州 730050

收稿日期:2024-09-19 发布日期:2025-12-31
通讯作者: 李策(1974-),男,辽宁营口人,博士,教授,博导.Email:lice@lut.edu.cn
基金资助:
国家自然科学基金(62363025),甘肃省高等学校创新基金(2021A-109)

A dual-branch global-local attention network for insulator defect detection

JIAO Liang¹, WANG Zhong-shun², WANG Si-run¹, LI Ce³

1. China Tianshui Great Wall Switch Factory Group Co. Ltd., Tianshui 741020, China;
2. School of Automation and Electrical Engineering, Lanzhou University of Technology, Lanzhou 730050, China;
3. School of Microelectronics Industry Education Integration, Lanzhou University of Technology, Lanzhou 730050, China

Received:2024-09-19 Published:2025-12-31

摘要/Abstract

摘要： 绝缘子的表面缺陷可能会严重影响其性能,甚至导致电力故障,因此对绝缘子表面缺陷的准确识别和精确定位是确保电力系统安全运行的关键.绝缘子缺陷检测的重点是准确识别和精确定位其材料表面缺陷,然而基于传统深度学习的方法在检测绝缘子微小缺陷时存在检出率差等局限性.针对该问题,提出了一种基于全局-局部双分支注意力模块(GLDA)的改进型YOLOX-S算法.首先,全局分支使用传统的注意力机制,并对K和V进行下采样以减少计算量,从而捕捉低频全局信息;同时,局部分支有效地融合了共享权重和上下文感知权重,以聚合高频的局部信息;其次,采用自适应空间特征融合(ASFF)模块提取缺陷的多尺度信息,提高绝缘子缺陷检测精度;最后,引入完整交并比(CIoU)损失指标对基线模型的训练损失进行优化.实验结果显示,该方法的准确率、召回率和平均精度均值分别达到93.2%、92.9%和95.2%.

关键词: 绝缘子缺陷检测, 深度学习, 注意力机制, 多尺度

Abstract: Surface defects of insulators can severely deteriorate their performance and may even lead to power failures. Therefore, accurate identification and precise localization of surface defects on insulators are essential for ensuring the safe operation of the power system. The focus of insulator defect detection is the accurate identification and localization of defects in its materials. However, conventional deep learning-based methods face limitations such as poor detection rates when identifying small defects. To address this issue, this paper proposes aglobal and local dual-branch attention module (GLDA). Firstly, the global branch utilizes a traditional attention mechanism but applies downsampling to K and V to reduce computational load, capturing low-frequency global information. The local branch effectively integrates shared weights and context-aware weights to aggregate high-frequency local information. Secondly, an Adaptive Spatial Feature Fusion (ASFF) module is used to extract multi-scale information of defects, enhancing detection accuracy. Lastly, the CIoU algorithm is introduced to optimize the training loss of the baseline model. Experimental results show that the proposed method achieves accuracy, recall, and mAP of 93.2%, 92.9%, and 95.2%, respectively.

Key words: insulator defect detection, deep learning, attention mechanism, multi-scale

中图分类号:

TP393.1

焦亮, 王宗顺, 王思润, 李策. 一种基于全局-局部双分支注意力网络的绝缘子缺陷检测算法[J]. 兰州理工大学学报, 2025, 51(6): 107-115.

JIAO Liang, WANG Zhong-shun, WANG Si-run, LI Ce. A dual-branch global-local attention network for insulator defect detection[J]. Journal of Lanzhou University of Technology, 2025, 51(6): 107-115.

参考文献

[1] ZHANG Z,ZHANG B,LAN Z,et al.FINet:an insulator dataset and detection benchmark based on synthetic fog and improved YOLOv5[J].IEEE Transactions on Instrumentation and Measurement,2022,71:1-8.
[2] HAO K,CHEN G,ZHAO L,et al.An insulator defect detection model in aerial images based on multiscale feature pyramid network[J].IEEE Transactions on Instrumentation and Measurement,2022,71:1-12.
[3] 张鹏林,姚吉,牛显明,等.基于射线照相的石墨电极缺陷检测与识别[J].兰州理工大学学报,2021,47(1):16-21.
[4] LIU X,MIAO X,JIANG H,et al.Data analysis in visual power line inspection:an in-depth review of deep learning for component detection and fault diagnosis[J].Annual Reviews in Control,2020,50:253-277.
[5] LI X,LI Z,WANG H,et al.Unmanned aerial vehicle for transmission line inspection:status,standardization,and perspectives[J].Frontiers in Energy Research,2021,9:713634.
[6] 亢洁,王勍,刘文波,等.融合CAT-BiFPN与注意力机制的航拍绝缘子多缺陷检测网络[J].高电压技术,2023,49(8):3361-3376.
[7] ZHAO W,XU M,CHENG X,et al.An insulator in transmission lines recognition and fault detection model based on improved faster RCNN[J].IEEE Transactions on Instrumentation and Measurement,2021,70:1-8.
[8] LUO J,LU L,DONG S.Detection of spontaneous explosion of transmission line insulators based on improved mask-RCNN algorithm[C]//Proceedings of the International Conference on Electrical,Control and Information Technology. Kunming,China:VDE,2022:1-5.
[9] HE K,GKIOXARI G,DOLLÁR P,et al.Mask R-CNN[C]//Proceedings of the IEEE International Conference on Computer Vision.Venice,Italy:IEEE Press,2017:2961-2969.
[10] GIRSHICK R.Fast R-CNN[C]//Proceedings of the IEEE International Conference on Computer Vision.Santiago,Chile:IEEE Press,2015:1440-1448.
[11] REN S,HE K,GIRSHICK R,et al.Faster R-CNN:towards real-time object detection with region proposal networks[J].IEEE Transactions on Pattern Analysis and Machine Intelligence,2017,39(6):1137-1149.
[12] REDMON J,DIVVALA S,GIRSHICK R,et al.You only look once:unified,real-time object detection[C]//Proceedings of the IEEE Conference on Computer Vision.Venice,Italy:IEEE Press,2017:779-788.
[13] ZHANG M,LI Q.Research on insulator detection algorithm of YOLOv5 fused with attention mechanism[C]//Proceedings of the International Conference on Artificial Intelligence,Information Processing and Cloud Computing. Kunming,China:IEEE Press,2022:384-388.
[14] YANG Z,XU Z,WANG Y.Bidirection-fusion-YOLOv3:an improved method for insulator defect detection using UAV image[J].IEEE Transactions on Instrumentation and Measurement,2022,71:1-8.
[15] HE M,QIN L,DENG X,et al.MFI-YOLO:multi-fault insulator detection based on an improved YOLOv8[J].IEEE Transactions on Power Delivery,2023,39:168-179.
[16] ZHU X,LYU S,WANG X,et al.TPH-YOLOv5:improved YOLOv5 based on transformer prediction head for object detection on drone-captured scenarios[C]//Proceedings of the International Conference on Computer Vision.QC,Canada:IEEE Press,2021:2778-2788.
[17] CHEN Y,LIU H,CHEN J,et al.Insu-YOLO:an insulator defect detection algorithm based on multiscale feature fusion[J].Electronics,2023,12(15):3210.
[18] 张剑锐,魏霞,张林鍹,等.改进YOLOv7的绝缘子检测与定位[J].计算机工程与应用,2024,60(4):183-191.
[19] SUN Q,HE N,ZHANG J,et al.Lightweight human pose estimation method based on non-local high-resolution network[J].Journal of Computer Applications,2022,42(5):227-234.
[20] 李策,胡治佳,肖利梅,等.基于全局和局部低秩矩阵分解的图像显著目标检测算法[J].兰州理工大学学报,2015,41(6):79-83.
[21] 胡海涛,杜昊晨,王素琴,等.改进YOLOX的药品泡罩铝铂表面缺陷检测方法[J].图学学报,2022,43(5):803-814.

一种基于全局-局部双分支注意力网络的绝缘子缺陷检测算法

A dual-branch global-local attention network for insulator defect detection

PDF

可视化

摘要/Abstract

引用本文

使用本文

参考文献

相关文章 15

编辑推荐

Metrics

本文评价

[1]	林玉萍, 刘梦皎, 王明豪, 张栋, 许美凤, 李策. 面向多模态皮肤病语料库的可变形分区注意力黑色素瘤识别方法[J]. 兰州理工大学学报, 2025, 51(5): 92-99.
[2]	张玺君, 苏晋, 陈宣, 尚继洋, 崔勇. 基于Attention-T-GRU的短时交通流预测[J]. 兰州理工大学学报, 2025, 51(4): 107-113.
[3]	陈辉, 胡荣海, 席磊. 基于长短期记忆神经网络局部认知结构的最优非线性卡尔曼滤波[J]. 兰州理工大学学报, 2025, 51(4): 88-94.
[4]	王江, 剡昌锋, 卢家伟, 王瑞民, 张永明. 融合多头自注意力机制的故障命名实体识别[J]. 兰州理工大学学报, 2025, 51(4): 33-42.
[5]	赵宏, 张陈鹏, 王奡隆, 张扬. 基于细粒度字词特征的中文作者识别模型[J]. 兰州理工大学学报, 2025, 51(3): 99-106.
[6]	汪小虎, 赵荣珍, 邓林峰, 郑玉巧. 基于多尺度空洞卷积神经网络的滚动轴承故障识别方法[J]. 兰州理工大学学报, 2025, 51(3): 55-63.
[7]	马宁, 赵荣珍, 郑玉巧. 基于CEEMDAN与改进一维多尺度卷积神经网络结合的滚动轴承故障诊断[J]. 兰州理工大学学报, 2025, 51(1): 45-54.
[8]	杜先君, 王紫阳. 电动汽车IGBT剩余使用寿命预测[J]. 兰州理工大学学报, 2024, 50(2): 77-86.
[9]	陈辉, 牛丽丽, 付辉, 张天佑, 席磊. 基于多尺度残差和注意力机制的图像去雾算法[J]. 兰州理工大学学报, 2024, 50(2): 69-76.
[10]	赵小强, 王涛. 基于边缘约束与范数比值的图像盲复原[J]. 兰州理工大学学报, 2024, 50(1): 76-83.
[11]	杜先君, 邱小彧. 基于改进LSTM神经网络的化工过程故障诊断[J]. 兰州理工大学学报, 2023, 49(6): 72-79.
[12]	王文婷, 刘姝君, 张耀聪, 杜小泽, 许潼. 基于Seq2Seq深度学习模型的焦炉煤气发生量预测方法研究[J]. 兰州理工大学学报, 2023, 49(5): 50-58.
[13]	王星, 晏榕璟. 深度可分卷积结合多通道注意力的垃圾图像快速分类模型[J]. 兰州理工大学学报, 2023, 49(3): 88-93.
[14]	王燕, 赵妮妮, 范林. 基于图神经网络特征交叉的协同过滤算法[J]. 兰州理工大学学报, 2023, 49(1): 94-102.
[15]	郭润兰, 史方青, 范雅琼, 何智. 基于卷积神经网络的壁面清洗机器人障碍物检测识别算法[J]. 兰州理工大学学报, 2022, 48(4): 83-89.